NBA Passing Network Analysis_2020-21 ~ 2024-25

Creator

C.T. Chin

This project was developed with AI assistance using Monica, with Claude selected as the primary language model.

Data Source

NBA API - nba_api

Data Processing

Data processed using R programming language

Data Description

This project analyzes NBA passing networks based on the NBA's tracking data from the Passes Dashboard. The data represents the total number of passes from one player to another during a season. Each edge in the network represents the cumulative number of passes between two players throughout a season.

How to Use This Tool

This tool is structured as follows:

• Data Sets: Choose between "All Passes" (complete passing data) or "Filtered Passes" (only players with 40+ games and edges with 50+ passes)
• Seasons: Select from available NBA seasons (2020-21 to 2024-25)
• Teams: Browse or search for specific NBA teams
• Visualizations: For each team, access network plots, community detection, and hierarchy analysis
• Network Metrics: View detailed network indicators for each team or entire seasons

Use the filters at the top of each dataset section to search for specific teams, seasons, or plot types. The "Expand All" and "Collapse All" buttons help navigate through the content.

Network Indicators Explained

Indicator	Description	Interpretation in Passing Networks
nodes	Number of vertices in the network	Total number of players in the team's rotation
edges	Number of connections in the network	Total number of unique passing connections between players
density	Proportion of potential connections that are actual connections	Higher values indicate more interconnected passing patterns; lower values suggest more specialized passing roles
vertex_connectivity	Minimum number of vertices needed to remove to disconnect the network	Higher values indicate more robust passing networks that can withstand player absences
edge_connectivity	Minimum number of edges needed to remove to disconnect the network	Higher values indicate more redundant passing pathways
num_component	Number of connected components	Ideally 1; more components suggest disconnected player groups
max_component_size	Size of the largest connected component	Closer to total nodes indicates more integrated team play
biconnected_components	Number of biconnected components	Fewer components suggest more robust passing structures
cohesive_blocks	Number of cohesive blocks in the network	More blocks indicate more hierarchical team structure
max_cohesion	Maximum cohesion value in the network	Higher values indicate stronger core passing groups
cent_indegree	Centralization of incoming connections	Higher values indicate team passing is focused toward specific players
cent_outdegree	Centralization of outgoing connections	Higher values indicate passing originates from specific players
cent_inclo	Centralization of incoming closeness	Higher values indicate some players are primary pass receivers
cent_outclo	Centralization of outgoing closeness	Higher values indicate some players are primary pass distributors
cent_betw	Betweenness centralization	Higher values indicate presence of key passing intermediaries
centr_eigen	Eigenvector centralization	Higher values indicate hierarchical influence in the passing network
walktrap_communities	Number of communities detected by Walktrap algorithm	More communities suggest distinct passing groups within team
walktrap_modularity	Modularity score for Walktrap communities	Higher values indicate stronger community structure
edge_betweenness_communities	Number of communities detected by Edge Betweenness algorithm	More communities suggest distinct passing groups based on critical passing paths
edge_betweenness_modularity	Modularity score for Edge Betweenness communities	Higher values indicate stronger community structure
louvain_communities	Number of communities detected by Louvain algorithm	More communities suggest distinct passing groups optimized for modularity
louvain_modularity	Modularity score for Louvain communities	Higher values indicate stronger community structure
best_modularity	Highest modularity score among all community detection methods	Higher values indicate stronger overall community structure
tau_rankedcluster	Ranked clustering coefficient	Measure of hierarchical structure in passing patterns
tau_cluster	Clustering coefficient tau	Measure of local clustering in passing patterns
tau_balance	Balance coefficient tau	Measure of balance in passing structure

Visualization Types

Network Plots: In these visualizations, circles represent players and lines represent passes between players. Larger circles indicate players with more seasons of experience, while thicker lines represent higher passing volume during the season.

Community Detection: These plots show different algorithms' attempts to identify natural groupings of players based on passing patterns. Different colors represent different detected communities. The three methods (Walktrap, Edge Betweenness, and Louvain) use different approaches to identify these communities.

Hierarchy Analysis: Cohesive Blocks: These visualizations represent the hierarchical structure of the team's passing network. They identify nested groups of players with increasing levels of interconnectedness. The blocks at higher levels represent more cohesive subgroups within the team, showing which players form the most tightly connected passing units.